The Research On Intrinsic Parameters In The Space Vision Camera Calibration

In the deep space exploration activities,the space vision camera is often used as a direct observation device to obtain geographic location information,which is closely relevant to the scientific task of topographic survey.CCD cameras are introduced into the aerospace industry because of theirs high-quality,small-sized and low-cost.Although they can meet the needs of two-dimensional image capture,these non-measuring cameras should be firstly calibrated in geometry,in order to achieve the destination of three-dimensional reconstruction such as the land restoration.Camera geometric calibration is based on the mapping between object points and image points to obtain higher precision camera intrinsic parameters(focal length,principal point).On the basis of the traditional calibration method,we develop a complete calibration algorithm to acquire more accurate camera intrinsic parameters by computing the images which are acquired by picturing a circular array calibration target of two dimensions in different positions.RMS is proved to be 0.178 pix in this study and can meet the engineering requirements(0.3pix).The main research contents are as follows:(1)Analyze and summarize the advantages and disadvantages of several typical calibration algorithms,and we propose an improved algorithm based on collinear constraint iterative solution.Firstly utilize the vanishing point estimation method to evaluate a more robust focal length parameter as an initial value,and then add the collinear constraint to eliminate the model distortion which the image noise brings.Secondly,we iteratively optimize the image point coordinates in the process of the collinear constraint.Finally,the camera intrinsic parameters are gotten by LM optimization.The modified algorithm overcomes the instability of the focal length calculation in Zhang's calibration.Moreover,it is more suitable for the actual imaging model.(2)In order to obtain the pixel coordinates corresponding with the circle centers in the object coordinate,the speckle removal method is employed to extract the pure feature circles.Subsequently,we locate centers of circles by the least squares curve and identify the target direction through the rectangular mark.Finally,we have access to the Delaunay triangulation for accomplishing the automatic sorting of the center coordinates correctly,and a complete image processing flow is formed,which reduces the manual recognition time and can automatically extract the target feature information.(3)An error correction method is utilized to further improve the camera calibration accuracy.After the rough points are removed based on the statistical method,the correction of the center asymmetry error is modified by the orthorectification.The experimental results show that the accuracy of the camera intrinsic parameters has been improved,which proves the effectiveness of the proposed method.